View
218
Download
0
Category
Preview:
Citation preview
7/24/2019 Drei Tae Gig Final
1/51
Gerold Schneider: Introduction to GB
Colloquium Gerold Schneider:
An Introduction to Government & Binding Theory
Colloquium Description: This colloquium is intended to give an introduction to one of
the major formal grammar theories: Government & Binding (GB), also known as
Principles & Parameters. It is a natural continuation and extension to the syntax taught
inIntroduction to Linguistics I andII. The colloquium rather aims to give an overview
than to dwell on technicalities. It will also discuss some fundamental problems of any
grammar theory and highlight similarities and differences to other approaches.
Chomskys theories about language knowledge and acquisition will not be discussed in
this course.
Course Requirements: The colloquium is situated at seminar level, i.e. you should
have passed your Linguistik-Akzessprfung, and it may be helpful if you have attended
a seminar beforehand.
Course Structure:We will deal with one topic in each meeting, i.e. seven topics in
total, as the chapters of these colloquium notes express. It will be a fully taught course,
there will be no student presentations, but discussions and questions are an essential
part of the colloquium. You are requested to read the corresponding chapter of these
notes before the colloquium (except for the first chapter, of course), which we will work
through carefully and discuss in class. I will recommend excerpts of additional
literature for most chapters. Please ask me if you would like to get copies of these
excerpts.
- 1 -
7/24/2019 Drei Tae Gig Final
2/51
Gerold Schneider: Introduction to GB
1 BACKGROUND OF GOVERNMENT & BINDING (GB)..........................................................3
1.1 LANGUAGESTRUCTURE...................................................................................................................41.1.1 Structure inside Structure.....................................................................................................4
1.1.2 Constituents............................................................................................................... ......... ..41.1.3 Phrase Structure or Rewrite Rules................................................................................ .......51.1.4 Decisions about Linguistic Structure...................................................................................51.1.5 Exloring the Elehant........................................................................................................!1.1.! Recursion.......................................................................................................................... ...."
1.2 COMPETENCEVERSUSPERFORMANCE............................................................................................ .71.2.1 State#ents on S$ntactic Per%or#ance..................................................................................&1.2.2 'sing Per%or#ance State#ents %or ()uto#atic* Parsing....................................................&
1.3 SYNTACTICRULESANDTHELEXICON.............................................................................................91.3.1 +erb Subcategorisation..................................................................................................... ....,1.3.2 Eli#inating Redundanc$....................................................................................................1-1.3.3 Lexicalis#........................................................................................................................ ...1-
1.4 TRANSFORMATIONS , DEEPANDSURFACESTRUCTURE..................................................................111.4.1 Dee and Sur%ace Structure................................................................................................11
1.4.2 he /0odel........................................................................................................................141.5 THELANGUAGEFACULTY............................................................................................................. 15
1.5.1 Princiles and Para#eters.................................................................................................151.5.2 Language )cuisition..................................................................................................... ....15
1.6 R ECOMMENDEDLITERATURE.........................................................................................................15
2 X-BAR SYNTAX.............................................................................................................................17
2.1 THEHEADOFAPHRASE................................................................................................................172.1.1 ) 'niersal Rewrite Rule........................................................................................... ........1&2.1.2 ransitie and ntransitie Preositions................................................................. ......... ..1&
2.2 INTERMEDIATECATEGORIES......................................................................................................... .22.2.1 irst Disadantage 6ne/Prono#inalisation.................................................................... .2-2.2.2 Second Disadantage )rgu#ents and )d7uncts................................................................21
2.3 THEX!"ARRULESCHEMATA........................................................................................................ 243 FUNCTIONAL CATEGORIES................................................................................................ .....27
3.1 SOMEDETECTIVE#OR$...............................................................................................................273.2 THEHEADOFASENTENCE........................................................................................................... 2%
3.2.1 8P ersus +P................................................................................................................ ......2&3.2.2 +/rile/9ar :.....................................................................................................................2&3.2.3 ;hat is S :..................................................................................................................... .....2,3.2.4 n%lection as a unctional Categor$..................................................................................3-3.2.5 8L as the
7/24/2019 Drei Tae Gig Final
3/51
Gerold Schneider: Introduction to GB
6.2 "AREPHRASESTRUCTURE............................................................................................................ 47
7 %AT %E AVE NOT COVERED............................................................................................4
BIBLIOGRA"Y...........................................................................................................................4'
1 Background of Government& Binding (GB)
RECOMMENDED READING for Chapter 1: Cook & Newson (1996: 1-49
GB has several predecessors and a successor is beginning to emerge now. While someparts of the theory have remained stable right from the start, others have developed
considerably. The most important predecessor is Transformational Grammar, which is
presented first in Chomsky (1957) and Chomsky (1965). As the name suggests,
transformations were and still are a most prominent part of the theory.
Transformations basically transform related structures into each other. Transformations
can e.g. express the relationship between the declarative sentence You are a studentand
the questionAre you a student ?. We will hear more about transformations in 1.4and
chapter4.
The successor that is emerging now is called The Minimalist Theory (Chomsky
1995), but it is not entirely clear yet in which direction it will develop. If we have
enough time I would like to deal with two suggestions made in Minimalism; Checking
Theoryin subchapter 6.1,andBare Phrase Structurein subchapter 6.2.
GB was first introduced in Chomsky (1981). Noam Chomsky, the inventor and
main protagonist of GB, is notorious for writing articles and books that are difficult to
read, and for inventing (and sometimes rebutting) a plethora of technical terms. While it
may certainly be interesting to read the original, be warned about the extremelydifficult style: Even experts admit they find it very difficult to follow! In a much milder
form, the same warning applies to any introductory book to GB: Do not be put off by
linguistic jargon and technical terms. Try to find their definition, you will usually
discover that a complicated name hides a more simple and fairly intuitive concept. It
has to be admitted, however, that many of the textbook discussions are rather technical,
which I will try to avoid here, whenever possible.
- 3 -
7/24/2019 Drei Tae Gig Final
4/51
Gerold Schneider: Introduction to GB
1.1 Language Structure
One of the main aims of linguistics is to find and describe the structure that is hidden
behind the utterances we make. Let us look at two examples:
( 0) Peter loves Mary.
( 0) *Loves Peter Mary.
We will intuitively accept ( 0) but reject ( 0). We will conclude from this, in
accordance with all linguists, that an English sentence usually has Subject-Verb-Object
(SVO) order. SVO, in this order, is the structure of an English sentence.
1.1.1 Structure inside Structure
If we see a sentence like
( 0) Peter loves a clever girl.
should we suggest that English has Subject-Verb-Article-Adjective-Object
structure?
Probably all linguists would still agree that ( 0) is also an example of the Subject-
Verb-Object structure of English, and that its object consists of several words, or that
the object has words connected to it on a secondary level, or something similar in
short, they would say that the object has a structure inside the structure of the sentence.
Language is generally supposed to have hierarchical structure, i.e. subordinate
structures within bigger structures.
1.1.2 Constituents
We can replace the single wordMaryby a sequence that consists ofwords, so-called
constituents, without rendering the sentence ungrammatical. Beside this well known
replacement testlinguists use a couple of additional tests to assess if a sequence of
words makes up a constituent (Radford 1988: 69 ff.). Constituents are bracketed:
Preposing: Only entire constituents can be preposed:I cant stand [your older
sister] [Your older sister], I cant stand.
Postposing: Only entire constituents can be postposed:He explained [all the
problems that he had encountered] to her.He explained to her [all the
problems that he had encountered].
- 4 -
7/24/2019 Drei Tae Gig Final
5/51
Gerold Schneider: Introduction to GB
Conjunction: Only constituents of the same level can be conjoined: He called
[Mary] and [a clever girl].
Short answers: Only constituents can be asked for or constitute a short
answer: [Who] did you call ? [A clever girl].
Pronominalisation: Only entire constituents can be replaced by pronouns:
Have you seen [the young student from Cambridge] ? Yes, I have seen
[him].
1.1.3 Phrase Structure or e!rite u"es
BothMaryand a clever girlare instances of a noun phrase (NP). The structure of a NP
can consist of a proper noun (PN) like Mary, or of an article (ART) followed by a noun
(N), which is in turn possibly followed by an adjective (ADJ). We can describe these
facts in so-called phrase structure rulesor rewrite rules, where can be read as
consists of (cf. e.g. Finegan (1994: 133))
( 0) NP PN
NP ART (ADJ) N
If we ask ourselves how the structure of the whole sentence (S) could be
expressed in this way, it will be natural to suggest the following rule, expressing the
discussed Subject-Verb-Object order (where V stands for Verb):
( 0) S NP V NP
We may find, however, that this rule predicts ungrammatical sentences like
( 0) *Peter laughs Mary.
Rule ( 0) would only allow transitive verbs. We have to formulate an additional
rule
( 0) S NP V
1.1.# $ecisions a%out Linguistic Structure
It has generally been accepted, however, that it is more elegant to distinguish between
sentences with transitiveand intransitiveverbs at a subordinate rewrite rule stage, not at
the S-level, the topmost level in the hierarchy of our small rewrite rule grammar.
Transitive and intransitive verbs are said to have too much in common to justify two
completely separate sentential representation rules. We therefore use a uniform rewrite
rule for the sentence by introducing a Verb Phrase (VP), which in turn contains a verb
in case of an intransitive verb or a verb followed by a noun for a transitive verb.
- 5 -
7/24/2019 Drei Tae Gig Final
6/51
Gerold Schneider: Introduction to GB
( 0) S NP VP
VP V
VP V NP
While this grammar is more elegant, as it allows a uniform treatment of all verbs,
the solution we have sketched with rules ( 0) and ( 0) cannot be said to be wrong in
any sense. We can at best find it less convincing and therefore discard it. We had to
take a decision as to how the underlying structure of a sentence could look like. While
we had good arguments for this decision and could possibly find more, nobody can be
forced to agree, because we cannot physically see the structure of a sentence under a
microscope. We will see later that there are cases where agreement among linguists is
more difficult to achieve, that different solutions are being discussed.
1.1. '"oring the '"ehant
Linguists are like people in a dark room who touch an elephant and try to describe it.
The one touching the tail will find it hairy, the one touching the trunk will find it soft
and skilful, the one touching a leg will describe it as rough-skinned and ponderous. Yet
none can know how the elephant, i.e. the structure of language, really looks like or if
it exists at all.
The conventional wisdom that a language like English has a real existence led to theidea that we only have to discover the structure of thelanguage. The question than
became:Howcould we discover its structure and the structure of language in general?
In later years it was thought, optimistically, that it was simply a problem of the blind
men, each examining a different part of an elephant. It was thought that more work was
needed on devising better test criteria and on how to interpret the results of the tests,
and that these results would eventually converge on a consistent overall description of
the elephant.
But, as one of my students gasped when realizing the implications of the domain
confusions, "There is no elephant!"
(Yngve 1996: 46)
Doubting the very existence of language structure is of course the most radical
formulation. By domain confusion, Yngve means that many linguists tend to present
their often debatable findings as if they were proven scientific facts, e.g. structures we
see under a microscope.
One of the characteristics of GB is indeed that everything is defined in terms of
structure-relations, so-called configurationally.
! Pri"ci#le $% eratio"s o" se"te"ces re'uire a"olege o+ the structural relatio"shi#s o+ the ors
- 6 -
7/24/2019 Drei Tae Gig Final
7/51
Gerold Schneider: Introduction to GB
We have seen that English has SVO order. Using the simple phrase structure
grammar just developed in ( 0) we get syntax trees like
SNP VP
,Peter V NP , , loves Mary
We can define subject to be the NP whose mother node is S, and objectto be the
NP whose mother is VP. GB in fact uses a further development of this simple
definition, which does not necessarily apply to all languages.
I do not want to discuss these issues in detail, but it is important to stress that the
assumptions that GB takes, the structures that GB suggests are often not be the only
solutions possible. On the other hand, a formal grammar like GB is consistent and
rigorous. Once a decision has been taken, it is followed consistently, and the assumed
decisions are often applied to decide related questions.
1.1.* ecursion
Let us return to our small rewrite grammar, which expresses one possible way to see
structure in a sentence. If we have a sentence like( 0) Sa"ra "os that Peter loves Mary
our small grammar fails to express it. The VP is far more complex in this
sentence. Instead of V+NP, as is usual for transitive verbs, it consists of V followed by
thatand the assertion of a fact that looks exactly like a sentence itself. We may indeed
suggest that the VP of ( 0) can be expressed by rule ( 0):
( 0) VP V -thatS
This allows us to analyse sentences like
( 0) e sus#ects that Sa"ra "os that Peter loves Mary.
If a rule (partly) rewrites as a previous rule or even as itself we speak of
recursion. Recursion can potentially go on forever, which allows, at least theoretically,
for endless structures. Recursion is one way in which, as it is said of language, we can
form an infinite number of structures from a finite number of rules.
1.2 Cometence versus Performance
- 7 -
7/24/2019 Drei Tae Gig Final
8/51
Gerold Schneider: Introduction to GB
The mental capacity that enables a speaker to utter well-formed sentences and to refute
ungrammatical ones is called competence. It can be described by a rule system similar
to (but of course much more complex than) the small rewrite grammar we have started
to write in 1.1. But such a grammar makes no statements about performance, i.e.
which sentences and structures are actually found in language use, and how frequent
they are. Although GB has developed in many ways from rewrite grammars, as we shall
see, it is still and intentionally a competence-based grammar theory. Chomsky uses the
terms I-language (internal language) for competence, and E-language (external
language) for performance.
A major aim of Corpus Linguistics is to make statements about language
performance. At the first sight, Corpus Linguistics and a formal grammar theory, such
as GB, have few common interests. But they can benefit from each other, precisely
because of their different approach, in order to complement each other.
1.2.1 Statements on S+ntactic Performance
If we syntactically analyse a corpus by using the rigorous and consistent methods of a
formal grammar such as GB, we can find out which structures and which rules or
principles are used more frequently, and which are very rare in language use. It usually
also happens even with very sophisticated grammars that some sentences cannot be
correctly analysed, which indicates that the grammar is not (yet) complete. Corpus data
helps to test a grammar, as well as the assumptions that a grammar makes.
1.2.2 ,sing Performance Statements for (-utomatic)Parsing
Conversely, a grammar theory can greatly profit from the syntactical performance
statements that we can obtain as just described. Most sentences are ambiguousin many
ways, although we are rarely aware of it. The short sentence
( 0) Ti/e +lies lie a" arro.
describes at least the following four facts:
( 0a) Ti/e ela#ses too 'uicly.
( 0) Short1live i"sects sy/#athise ith a" arro.
( 0c) 2se your sto# atch to /easure the s#ee o+ i"sects3as i+ they ere a" arro.
( 0) 2se your sto# atch to /easure the s#ee o+ i"sects3as i+ you yoursel+ ere a" arro.
- 8 -
7/24/2019 Drei Tae Gig Final
9/51
Gerold Schneider: Introduction to GB
A very good automatic analyser would report all these readings and fail to decide
which could be best. But if a corpus provides us with morphosyntactic performance
data, in this case the frequency of the words in the sentence in their different possible
lexical categories, reading ( 0a) will be recognised as the most likely reading.
1.3 S+ntactic u"es and the Leicon
1.3.1 er% Su%categorisation
In a rewrite grammar fragment we might find the following rules
( 0)VP V
VP V NPThere are different verb subcategories, e.g. transitive and intransitive. How can
we make sure that intransitive verbs will only pick the former VP rule, and transitive
verbs only the latter rule? One solution is to add an argument to the rule, as follows:
( 0) VP V(i"tra"s)
VP V(tra"s) NP
The information on the transitivity of a verb depends on the verb itself. There is
no way to predict if, say an unknown verb in a foreign language we see in isolation,could be transitive or intransitive. This information is idiosyncratic for each verb, it
therefore needs to come from the lexicon. Lexical entries can be made as rewrite rules,
which would look as follows for the verbs to loveand to laugh:
( 0) V(i"tra"s) laugh
V(tra"s) love
There are many different types of verbs, however, which select all kinds of
complements. We have seen in 1.1.6 that the verb knowtakes a subordinating element
(so-called complementizer), i.e. that, followed by a sentence.
( 0) Sa"ra "os that Peter loves Mary
The verb asktakes a similar, but slightly different complement:
( 0) *Sa"ra ass that Peter loves Mary
( 0) Sa"ra ass i+ Peter loves Mary
But the verb wantonly takes a non-finite clause, a sentence without inflection
for GB as we shall see later.
- 9 -
7/24/2019 Drei Tae Gig Final
10/51
Gerold Schneider: Introduction to GB
( 0) *Sa"ra a"ts that Peter loves Mary
( 0) Sa"ra a"ts Peter to love Mary
Or, one more example, the verbputis ditransitive, but it requires a preposition,
usually onor intofor its second argument.
( 0) *Sa"ra #uts the oo
( 0) Sa"ra #uts the oo o" the tale
This diversity of verb requirements has led linguists to directly mark in the
lexicon what each verb takes as complement. The (simplified) lexical entries for some
of the above verbs look as follows in GB:
( 0) love Ver3 45 NP6
"o Ver3 45 that1S6
as Ver3 45 i+1S6
#ut Ver3 45 NP PP]
The underscore (_) refers to the subject, on which the verb usually places no
categorial restrictions. The fact that verbs are grouped into different subcategories is
called subcategorisation. When subcategorisation is expressed directly in the lexical
entry, as in ( 0), we speak of direct or strict subcategorisation. The bracketed
subcategorisations, e.g. [_ NP] are also referred to as subcategorisation frame, and the
verb lovesubcategorises foran NP.
1.3.2 '"iminating edundanc+
Strict subcategorisation, as explained above, expresses information that is already
contained in the phrase structure rules, it only helps to select the appropriate rule. It is
indeed questionable if it is useful to store the same information in two places, to have
redundant information. Because subcategorisation information depends on the
individual verb we cannot eliminate it from the lexicon, but we can eliminate it in the
phrase structure rule. We only have to make sure that lexical information is properly
inserted into the remaining partial phrase structure rules. The so-called Projection
Principleensures this (cf. Cook & Newson 1996: 20)
! Pri"ci#le 7 (Pro8ectio" Pri"ci#le)% The #ro#erties o+le9ical e"tries #ro8ect o"to the sy"ta9 o+ the se"te"ce.
1.3.3 Leica"ism
Not only verbs, but also many lexical entries show some idiosyncratic behaviour, i.e.
they specifically select certain other categories or lexical items. The noun relation, e.g.,
- 10 -
7/24/2019 Drei Tae Gig Final
11/51
Gerold Schneider: Introduction to GB
sometimes occurs alone, but it often subcategorises for a PP with the preposition to, or
two nouns connected by between:
( 0) a. The ver has a relatio" to the su8ect.
.*The ver has a relatio".
( 0) a. The i#lo/atic relatio"s etee" Turey a" :ra"are 'uite a.
. The i#lo/atic relatio"s are 'uite a.
Such information belongs into the lexicon. Most formal grammars, including GB,
have increasingly come to depend on the lexicon. The only remnant of phrase structure
rules left in GB are a structural constraint called X-bar, which will be presented in
chapter2. Grammars that rely mainly on the lexicon are called lexicalist.
In parsing technology, which is a topic in computational linguistics or artificial
intelligence, one distinguishes between top-downand bottom-uptechniques. Top-down
algorithms start at the topmost node, usually S, and try to apply the rewrite rules until
they reach the lexical items, like a phrase structure rewrite grammar. Bottom-up
algorithms start with the lexical items and build up the structure from them, just as a
lexicalist grammar does.
1.# /ransformations0 $ee and Surface
StructureTransformations, also called movements, are one of the first and most prominent parts
of GB and its predecessor, Transformational Grammar (Chomsky(1957)). The basic
idea of transformation is that functionally related sentences should be derived from the
same structure. E.g. the question
( 0) ;hat are you seei"g at the ci"e/a at the ci"e/a.
1.#.1 $ee and Surface Structure
GB distinguishes a surface structure(S-structure), the sentences as we see them
in print, and an underlying deep structure(D-structure). Both ( 0) and ( 0) are surface
structures, but they have an identical deep structure except for the question word what
instead of the movie title, of course. The surface structure of the declarative sentence
( 0) is (almost) identical to its deep structure. The deep structure of ( 0) is probably (but
we will revise this slightly in a minute):
- 11 -
7/24/2019 Drei Tae Gig Final
12/51
Gerold Schneider: Introduction to GB
( 0D) =ou are seei"g hat at the ci"e/a at theci"e/a +'(+
**+' >'-' I ' +'* '*8
/*)+ )8 8*)* +:*. I 8( )(+ ?)(> G" >* *)(' :*+ +( @(;+ ;(/ +'
=((?.
ABBMAN3 L:L:ANB3 $HH@. Intro!ction to Government %
&ining Theory3 7" e. &9+or% !lacell. T'* /(+
-(/@*'*)'-' * /(* ( * +?*) ;(
)+*8 +(8:, *.. 8*)+;:) -()++*)+ ( >': ()* )**8 )+*/*8+*
-+*(*, * 8-*8 + +'* *)+' +'*: 8**
7/24/2019 Drei Tae Gig Final
18/51
Gerold Schneider: Introduction to GB
2 5%ar S+ntaRECOMMENDED READING for Chapter !: Cowper (199!: 19-""# Ra$for$ (19%%:
16-1%
So far we have employed phrase structure rewrite rules to make a small English
grammar. We have heard in 1.3 that a part of them may be replaced by lexical
requirements projected up to the syntax, but this does not change anything about how
the structure of a sentence will eventually look.
2.1 /he 6ead of a Phrase
A typical pre-GB phrase structure rule grammar (taken from Cowper (1992: 20), which
I will follow for much of this subchapter) may have looked as follows (elements in
brackets are optional):
( 0) S NP VP
VP V (NP) (ADVP PP)
NP (DBT) (ADJP) N PP P NP
ADJP (DB) ADJ
Except for the categories S and PP, all the rules have only one obligatory element,
which is the word that gives the phrase its name, N for NP, V for VP, etc. It is
obviously the most prominent element in the phrase. It is therefore called the headof
the phrase. We will discuss if the NP needs to be compulsory in PPs below, and we will
discuss the special status of S in chapter 3.
The small grammar ( 0) allows us to analyse sentences like ( 0); heads are inboldface in the syntactic tree:
( 0) A rather large og le+t a o"e o" the car#et
S
NP VP
DET ADP N V NP PP ! og le+t
DEG ADJ DET N P NP r!"her large ! #one o"
DET N "he car#et
- 18 -
7/24/2019 Drei Tae Gig Final
19/51
Gerold Schneider: Introduction to GB
In addition to generally being obligatory, the head of a category has other
privileged characteristics. The semantic, syntactic and morphological properties of the
head usually become the properties of the whole phrase. If we consider the following
examples
( 0) a. the tall /a"
. the ha##y o/a"
c. the ro" og
. the i"tellige"t og
( 0) a. the ro" og
. the re car
c. the hite sugar
. the lue sy
all nouns in ( 0) refer to animate nouns, therefore the whole NP can be said to be
animate. On the other hand, although all the adjectives in ( 0) are colour adjectives, the
NPs can hardly be said to be colour NPs or something similar.
2.1.1 - ,niversa" e!rite u"e
Except for the category S, every rewrite rule of grammar ( 0) states that a phrase
consists of at least an element of the same category as the head, and possibly more. The
fact that every phrase is headed by a category of the same type as the whole phrase is
one of the cornerstones of X-bar theory:
! Pri"ci#le H (K1ar)% a #hrase alays co"tai"s a hea o+the sa/e ty#e.
We can suggest a universal rewrite rule, some kind of a skeleton for all rewrite
rules, a first version of X-bar theory. X stands for any category, e.g. N or V or
whatever.
( 0) KP ... K ... 4here ... co"tai"s co/#le/e"ts etc.6
Rule skeleton ( 0) will still show considerable variation in the individual rewrite
rules that will be fashioned for the grammar. We have seen that, except for PP, all non-
head elements are optional. Let us take a closer look at PPs:
2.1.2 /ransitive and 7ntransitive Preositions
It is obvious that many prepositions require a noun complement, as the following
ungrammatical sentences illustrate:
( 0) *She ra" i"to.
( 0) *e cut the rea ith.
- 19 -
7/24/2019 Drei Tae Gig Final
20/51
Gerold Schneider: Introduction to GB
However, there is a category of words that are very closely related to
prepositions, the so-called verbal particles, which do not require prepositions.
( 0) a. e tur"e the la/# o".
. She #ut the oo o".
c. Sue #ut her hat o".
Most of the words that can be verbal particles also belong to the lexical category
of prepositions. It has been suggested that verbal particles are indeed prepositions,
although prepositions that do not take complements. Verbs without complements are
called intransitive verbs. It is therefore natural to suggest that there are intransitive
prepositions(formerly called verbal particles) and transitive prepositions (the former
classical prepositions). As with verbs, some prepositions are always transitive, others
always intransitive, many can be both:
( 0) :"tra"sitive o"ly%
a. She thre the oo aay.
. *She thre the oo aay the i"o.
( 0) Tra"sitive o"ly%
a. *She ra" i"to.
. She ra" i"to the house.
( 0) !oth tra"sitive a" i"tra"sitive%
a. Sue #ut her hat o".
. Sue #ut her hat o" the shel+.
If we accept the hypothesis that verbal particles are prepositions1, we can also use
a phrase structure rule in which the NP is optional, and the complement requirements
for prepositions will be listed in the lexical entries of each preposition, as we have
suggested for verbs in 1.3. All the complements are now projected up from the lexicon,
the only non-head elements left in the phrase structure rules are entirely optional, they
only refer to optional extensions, modifiers and adjuncts. The universal rewrite rule
from ( 0) can now be reformulated:
( 0) KP (...) K (...) 4here all (...) are o#tio"al or#ro8ecte u# +ro/ the le9ico"6
1See Cowper (1992: 23 ff) for additional argumets, but also counter-arguments to this
hypothesis.
- 20 -
7/24/2019 Drei Tae Gig Final
21/51
Gerold Schneider: Introduction to GB
2.2 7ntermediate Categories
The phrase categories we have used until now have two levels of structure: the level of
the entire phrase (XP), and the level consisting of the head (X), complements projected
up from the lexicon and optional elements. Let us take a look at the representation for
the following sentence:
( 0) the you"g stue"t o+ #hysics +ro/ Fa/rige.
NP
DBT ADJ N PP PP
, , ,the you"g , P NP P NP
, , , , , , o+ N +ro/ N stue"t , , #hysics Fa/rige
The NP splits up many times at the same level, without any possible distinction
between the different elements. This is a so-called flat structure. It has two
disadvantages, however, as I shall explain now.
2.2.1 irst $isadvantage8 OnePronomina"isation
We have seen in 1.1.2 that one of the tests used to find constituents is the
pronominalisation test. Only entire constituents can be pronominalised.
( 0) The you"g stue"t
NP
DET AD N
"he $o%n& '"%den"
In( 0), the only constituent, therefore the only candidate for pronominalisation, is
the NP.( 0) : lie the 4you"g stue"t6.
: lie 4hi/6.
*: lie the 4hi/6.
*: lie the you"g 4hi/6.
There is one pronoun, however, which allows to replace these elements the
pronoun one:
( 0) : lie the 4you"g stue"t6.
: lie 4o"e6 as ell. =ou /ea" this 4o"e6
- 21 -
7/24/2019 Drei Tae Gig Final
22/51
Gerold Schneider: Introduction to GB
: #re+er the ol 4o"e6.
Either we have to give up the pronominalisation test, or we are forced to admit
that the pronominalised elements are also constituents. In the representation ( 0) they
are not. GB has chosen to keep the pronominalisation test and to assert that
pronominalised elements in ( 0) are indeed constituents. The pronominalised elements
are [student], [young student] and [the young student]. We know that [the young
student]is an NP, but [student] and[young student]have to be something between N
and NP. We have three N-levels now: We will call them N0or N, which is the lexical
category, N1 or N, which is the new intermediate level, and N2 or N, which is the
former XP.
( 0) the you"g stue"t
N> NP % N> 4the you"g stue"t6
DET N( : N( ) *$o%n& '"%den"+
"he
AD N( : N( ) *'"%den"+
$o%n& ,
N
'"%den"
As we can see, the intermediate category N is allowed to repeat itself several
times, e.g. when several adjectives precede a noun. We have seen in 1.1.6that this is
called recursion. The recursion of N in ( 0) corresponds to the following rewrite
grammar rule:
N ADJ N
There are some notational variants for N and N. The original notation was to
write a bar, or two bars respectively, on top of the N, or whichever other category.
Again, the universal phrase category label is X. The pronunciation for X is still X-bar,
for X it is X-double-bar hence the name X-bar syntax.
2.2.2 Second $isadvantage8 -rguments and -d9uncts
The two PPs at the end of the NP in ( 0) (repeated below for clarity) are both at the
same level, so they cannot be distinguished.
NP
DBT ADJ N PP PP , , ,
the you"g , P NP P NP , , , , , , o+ N +ro/ N stue"t , , #hysics Fa/rige
- 22 -
7/24/2019 Drei Tae Gig Final
23/51
Gerold Schneider: Introduction to GB
If we adopt the new notation, as suggested below, they are at different levels,
which means that they can be one-pronominalised. But as they both have an N as
mother node and sister node they can still not be distinguished from each other:
N> NPDET N(
"he
AD N(
$o%n&
N( PP
N( PP P NP
, , ,
N P NP ro. N(
'"%den" o , ,
N( N
, /!.#rid&e
N
h$'ic'
We should expect, therefore, that they have equal status, that they can be deleted
and replaced at will.
( 0) a. the you"g stue"t
. the you"g stue"t o+ #hysics
c. the you"g stue"t +ro/ Fa/rige
. the you"g stue"t o+ #hysics +ro/ Fa/rige
e.*the you"g stue"t +ro/ Fa/rige o+ #hysics
For some reason, the PP of physicshas to stay closer to the noun than the PPfrom
Cambridge. We have seen in 1.3.3that some nouns tend to subcategorise for PPs, as if
they were transitive. They are called relational nouns. In all of them, the subcategorised
PP has to occur closer to the noun than additional PPs. Here are some examples for the
nouns chairman,pieceand relation.
( 0) a. the chair/a" o+ the oar i" the lue suit
.*the chair/a" i" the lue suit o+ the oar
( 0) a. the #iece o+ cae o" a #late
.*the #iece o" a #late o+ cae
( 0) a. the relatio" o+ vers to "ou"s i" our iscussio"
.*the relatio" i" our iscussio" o+ vers to "ou"s
A chairman always has to be a chairman of something, a piece is a piece of
something, a relation is always a relation of something to something or a relation
between two things, much as whenever you kick, you kick something, or when you
love, you love somebody transitive verbs, transitive prepositions and relational nouns
- 23 -
7/24/2019 Drei Tae Gig Final
24/51
Gerold Schneider: Introduction to GB
all subcategorise for complements. These subcategorised elements are also called
arguments. They are usually obligatory, but can be dropped in many cases:
( 0) a. e has s#e"t hours ici"g (the all)
. this relatio" (to "ou"s) is a i++icult issue
c. Lovi"g (so/eoy) sets you +ree.
The optional additional elements which are neither heads nor complements (all
complements are projected up from the lexicon) are called adjuncts. They can be
modifiers, PPs etc. They have much less influence on the semantics of the head than
arguments, and they are always further away from the head than arguments, as we have
seen for nouns in ( 0). An example for verbs:
( 0) a. She e#e"s o" her +rie"s +or goo co/#a"y..*She e#e"s +or goo co/#a"y o" her +rie"s.
How can we distinguish between arguments and adjuncts in X-bar theory ? The
X-bar suggestion is that adjuncts are attached under recursion to X,i.e. with X both as
a mother node and sister node, while arguments are attached when X is rewritten to
X0:
( 0) K1ar A8u"ct Rule% K = K *here i' "he!d%nc"+
( 0) K1ar Argu/e"t Rule% K (=)(O) K *here !nd !re"he !r&%.en"' !' '%#c!"e&ori'ed or #$ "he leic!l en"r$
o "he he!d !nd roec"ed % "o "he '$n"!+
When we apply these rules, we get the following representation for ( 0) the young
student of physics from Cambridge.(with the crucial differences in boldface):
N> NP
DET N(
"he
AD N(
$o%n&
N PP
N PP P NP stue"t , , P NP ro. N(
o , ,
N( N
, /!.#rid&e
N
h$'ic'
- 24 -
7/24/2019 Drei Tae Gig Final
25/51
Gerold Schneider: Introduction to GB
While recursion, and therefore the admission of new adjuncts, may go on forever,
argument positions may only be occupied once. We can easily imagine sentences with
many adjective and PP adjuncts:
( 0) The you"g ha"so/e clever right stue"t +ro/
Fa/rige ith a rai"coat sta"i"g u"er the #orch
But a recursion of arguments, i.e. complements, is impossible, or it conveys a
different fact from what we intended:
( 0) a. *She loves Peter Mary.
. *Fats live o" /eat o" /il.
c. *A stue"t o+ #hysics o+ /aths
. *The relatio" o+ vers o+ a8ectives to "ou"s
An attentive reader may wonder why the adjunct rule ( 0) has an obligatoryadjunct Y, instead of bracketed and hence optional one. The problem would be that
then the rule could enter endless recursion without ever taking up an adjunct. Because
the rule is recursive it is the rather rule itself which is optional, instead of its elements.
Recursion will only take place if more adjuncts need to be placed in the syntactic
structure. In an adjunct-free sentence, the adjunct rule will never apply.
2.3 /he 5%ar u"e Schemata
For every phrase X, only following three rewrite schemata, skeletons of former rewrite
rules, remain:
! Pri"ci#le $0(K1ar)% K1ar Rule1sche/ata%
S#eci+ier Rule% K (S#eci+ier) K
tio"al A8u"ct Rule% K A8u"ct K
Argu/e"t Rule% K (Arg$) (Arg7) K
Specifiers are articles in NPs, in ADJPs or VPs they may be adverbs. Nothing hasbeen said about the orderof the constituents in the rules. It is assumed that this should
be a language-dependent parameter. In English, e.g., specifiers come before X, but
arguments after X. An English X-bar phrase can e.g. take one of the following shapes:
o"e a8u"ctio"3 o"e argu/e"t%
K>Sec (
Ad%nc" (
- 25 -
7/24/2019 Drei Tae Gig Final
26/51
Gerold Schneider: Introduction to GB
0 Ar&1
"o a8u"ctio"3 "o argu/e"t%
K>
Sec ( ,
0
"o a8u"ctio"3 o"e argu/e"t%
K>Sec (
0 Ar&1
to a8u"ctio"s3 to argu/e"ts%
K>Sec (
Ad%nc" (
Ad%nc" (
0 Ar&1 Ar&2
e"c e"c
Adjuncts are inserted by recursion of the adjunct rule, as required by the sentence
to be analysed. Arguments are present or not as required by the lexical entry of the
head. Arguments in turn are usually phrases, the head of the phrase introduced by this
argument has new arguments, etc. In this fashion, the representation for an entire
sentence is built up. X-bar depends heavily on the lexicon and the projection principle
we have met in 1.3.2. X are also called maximal projectionsto indicate the lexicalist
character of GB. The rule schemata that are left in GB cannot be called rules any
longer, they are too schematic, and their whole purpose it to direct the lexical
projections. They restrict the possible lexical projections. For this reason, X-bar is also
called a structural constraint. X-bar is the only structural constraint on the syntax in
GB, everything else is lexical projection.
When not needed, the intermediate level is often eliminated in syntactic
representations for the sake of legibility, and the terms XP and X are used
interchangeably.
Specifier nodes often remain empty, and as we shall see, they often make good
landing sites for elements under movement. We shall see in chapter 4that mainly heads
- 26 -
7/24/2019 Drei Tae Gig Final
27/51
Gerold Schneider: Introduction to GB
and specifiers can move. We will also see that specifiers move to specifier positions,
and heads to head positions.
- 27 -
7/24/2019 Drei Tae Gig Final
28/51
Gerold Schneider: Introduction to GB
3 unctiona" Categories
3.1 Some $etective :ork
NO READING 'GGE')ION' for '*+,hapter "1
At the current stage of our progress in understanding GB, we have a couple of
mysteries and indications which we would like to solve:
In 1.4.1we have seen that movement will always be structure-preserving, i.e.
that D-structure (deep structure) and S-structure (surface structure) of a given
sentence are the same, except that movement has taken place in the surface
structure, as in ( 0), repeated for clarity:
( 0S) ;hat$are7you t2seei"g t1at the ci"e/a (): +'+
): *)+*)-* ' +( =* ;)+* +'+ ;)+*)*, .*. );*-+() ;( +*)*, @*-+, @*(),
)/=* *+-. +'* +(-'+()* (; *)+*)-*'((8.
F( +'(* >'( * + )-()
7/24/2019 Drei Tae Gig Final
33/51
Gerold Schneider: Introduction to GB
+-+*, +' );(/+() >(8 )(+ =* --*=* +( +'*
NP (
Pe"er
V;
-'
V(
V ;
!n"
NP (
@!r$
V;
"o ,
V
&o
We shall see in chapter 4that want actually moves up to I (and remains there) in
order to take up the inflectional affix and to assure the usual verb-second position.
3.2.* /he Su%9ect under the 7P assumtion
I; >* --*@+ +'* *+() +'+ IP +'* )*> +(@/(+ @'* (; *'** -) >* @-* +'* =*-+ NP, ;(/*: +'* +* (; VP ) +'* (8 S NP VPrule?
First we need not forget that we only want to speak about the D-structure place of
the subject NP, because, as we shall see in chapter 4, it may move to other positions.
Until few years ago, subject NPs were suggested to reside in the empty specifier
position of IP, as just seen in ( 0):
:>
NP (
0 VP
- 33 -
7/24/2019 Drei Tae Gig Final
34/51
Gerold Schneider: Introduction to GB
More recently, however, the subject NP has been suggested to originate from
inside the VP, from where it moves up to I-Specifier even in most declarative
sentences.
:>
(
0 VP
NP V(
This hypothesis is known as the subject-within-VP hypothesiswhich we shall see
in4.
3.3 Com"ementi>ers
NO READING RECOMMENDA)ION' for '*+,hapter ""
After allowing the functional category I to be a head, many other categories followed.
Complementizers are words like that, ifor whether, which introduce finite subordinate
clauses. A complementizer phrase (CP) introduces an IP, it is therefore a new topmost
phrase. Let us look at an example sentence all unnecessary intermediate levels are
eliminated, and it is the D-structure representation:( 0) : ho#e that Mary ill co/e.
/P
P
NP VP
,
V(
V /P
7/24/2019 Drei Tae Gig Final
35/51
Gerold Schneider: Introduction to GB
from CP to IP to VP is not prescribed by X-bar, it follows entirely from the lexicon. A
complementizer (even an empty one as in ( 0)) naturally subcategorises for an IP. An
inflection I naturally subcategorises for a verb. A verb has more idiosyncratic
subcategorisational requirements, as discussed. This is the stage of the theory which I
will mainly use.
3.# ?ther unctiona" Categories
NO READING RECOMMENDA)ION' for '*+,hapter "4
Recently, many linguists have started to use DP (determiner phrase), which heads and
subcategorises for an NP. There have also been suggestions to split up IP into several
functional phrases. IP marks both tense, and agreement to subject, and agreement to
object, so that e.g. splits into tense-phrase, agreement-subject-phrase and agreement-
object-phrase have been suggested. I did not want to discuss these considerably
technical issues in an introductory course, however.
It is indeed questionable if functional categories, many of them bordering on the
line between syntax and morphology across languages, should be heads. Some linguists
prefer to treat them as so-called markers, which are headed by the lexical category they
mark. The authors of a rival theory to GB named Head-Driven Phrase Structure
Grammar (HPSG) criticise that many of the arguments brought forth in support of
functional heads depend on GB-internal assumptions (Pollard & Sag 1994: 44) like
X-bar theory.
On our account, a marker is a word that is functional or grammatical as opposed to
substantive, in the sense that its semantic content is purely logical in nature (perhaps
even vacuous). A marker, so-called because it marksthe constituent in which it occurs,
combines with another element that heads that constituent. In addition to the
complementizers that and for, other examples of markers include the comparative
words thanand as, the case-marking post-clitics of Japanese and Korean, and perhaps
nonpredicative adpositions in (the vast majority of) languages where adpositions
stranding does not occur.
(Pollard & Sag 1994. 44-5)
- 35 -
7/24/2019 Drei Tae Gig Final
36/51
Gerold Schneider: Introduction to GB
# ovement and Case /heor+
#.1 - $etai"ed 'am"e of ovement
RECOMMENDED READING for '*+,hapter 41: Ra$for$ (199: !16-!".
In the introduction to movement in 1.4I have shown how the S-structure of a question
is derived from its D-structure, the latter being almost identical to both the D-structure
and the S-structure of the corresponding declarative sentence. I have said that GB
postulates that movement is structure-preserving, i.e. that the positions to which
elements can move have to be provided for at D-structure (I have called themE1and
E2 for convenience, but this is not GB terminology). Let consider the examples again,
the D-structure ( 0D) and its S-structure ( 0S):
( 0D) E1 E2=ou are seei"g hat at the ci"e/a
Recommended